The present inventive subject matter relates generally to the art of image analysis. Particular but not exclusive relevance is found in connection with the analysis of images captured by red light and/or other traffic enforcement cameras. Accordingly, the present specification makes specific reference thereto. It is to be appreciated however that aspects of the present inventive subject matter are also equally amenable to other like applications.
For images captured with red light, traffic enforcement and/or other like automated and/or unattended cameras it is often desirable to know when or under what conditions an image was captured. For example, it may be desirable to classify images of a given scene (e.g., such as a roadway or traffic intersection) as being captured during the daytime (i.e., generally under relatively high natural ambient lighting conditions), or as being captured during the nighttime (i.e., generally under relatively low natural ambient lighting conditions). However, such classification of images can be complicated by scene illumination from various sources, e.g., such as artificially illuminated billboards and signs, traffic signals and lights, vehicle headlights, etc. Furthermore, difficulties can be experienced when attempting to distinguish daytime images captured with relatively low natural ambient lighting (e.g., images captures at dusk or dawn or with overcast conditions) from relatively brightly lit nighttime images (e.g., due to artificial lighting from various sources).
Accordingly, a new and/or improved method, system and/or apparatus for classifying images is disclosed which addresses the above-referenced problem(s) and/or others.